Chapter 4 System Design
4.2 Phase Demand Determination
4.2.4 Phase Demand Function
Now, we define the phase demand function according to the above mentioned passenger waiting time, bus schedule delay ratio and bus headway deviation ratio. The phase demand function is defined as follows:
21
At each intersection, if there is an overtime phase, the overtime phase will be selected.
Otherwise, the phase f with the greatest value will be assigned as the next phase.
22
Chapter 5 Simulation
In this section, we evaluate the performance of ATCB by using NetLogo simulator [23] (version 4.1.3). We compare ATCB with the traditional predefined fixed-time scheme with bus priority strategy like TDTSP [19] and an adaptive fuzzy logic control (AFLC) [21].Beside bus priority, we also compare ATCB with an actuated traffic control and a non-fixed sequence control scheme [10]. The details of each scheme are described below.
We modify the TDTSP: We use a fixed sequence traffic signal control scheme and we benefit bus by extending the current phase if there is a bus can pass through the intersection by the current phase. If there are two buses on different routes meet in an intersection, we compare the headway deviation and schedule delay to decide which bus will be benefits.
Then, we modify AFLC: Like TDTSP, we also adopt a fixed sequence traffic signal control scheme and decide whether to extend the current phase or switch to next phase.
In our modification, we compare ordinary vehicles and buses of the current phase with ordinary vehicles and buses of the next phase to decide whether to extend the current phase or switch to the next phase.
Actuated traffic control method controls signals by detecting the coming vehicles.
It places sensors at a short distance near the intersection, if the sensor find there are
23
vehicles will cross the intersection in a short period, it will extend the current phase until reach its maximum green time.
In a non-fixed sequence scheme, when the current phase is going to end, it will find the most suitable phase from all phases, the original scheme consider many factors, in our modification, we only focus on the number of vehicles, the phase has the biggest number of vehicles will be selected as the next phase.
We analyze the simulation results of total waiting time of vehicles, total waiting time of buses, total passengers’ waiting time, average bus schedule deviation and average bus headway deviation.
5.1 Simulation Environment
As shown in Figure 10, we perform the simulation on a network of 8×8 traditional four-direction intersections, and the length of roads is 500 meters. The length of the waiting area on each road is 200 meters, and each road has four lanes, two are approaching lanes, and two are leaving lanes. We generate the ordinary vehicles on the edge roads of the network in a rate of 10 vehicles/minutes. Each vehicle are created with a speed of 14m/s. The acceleration of vehicles is assigned as 2m/s2, it means that each vehicle will reach its limit speed in 7 second. The deceleration of vehicles is 4m/ s2. Each vehicle will keep a safe distance when it is driven. And we adopt each vehicle carry average two passengers. In this network, we set five bus routes (RouteA, RouteB, RouteC, RouteD and RouteE), each bus on different bus routes enter this map with different frequencies (predefined headway). And we let them meet at an intersection to generate a pivot intersection, the bus routes is shown in Figures 10. And passengers on each bus is assigned in a range from 10~20. We run each simulation in 2 hours. The
24
detail parameter of simulation is shown in TABLE 1.
We generate these five routes randomly. First, we randomly select a pivot intersection, and the distance between the pivot intersection and edges of this map is at least two intersections. For each route, we give an entry and an exit randomly, and the entry can’t be also the exit. The buses of the route will pass through the pivot intersection, if the route is illegal or the length of this route is less than eight intersections, we will generate a new route until the route is legal and enough long. The frequency of each bus route is from 3 minutes to 10 minutes randomly.
Route A Route B
Route C Route D
Route C
Route E
Figure 10. An Example of random network and bus routes
25
Table 1. Simulation parameters
Map size 8*8 grid
Length of roads 500 m
Length of waiting area 200 m
Traffic flow the edge road 10 vehicles/min
Speed limit 14 m/s (50 km/h)
Acceleration 2 m/s2
deceleration 4 m/s2
Passengers of a ordinary vehicle
2
Passengers of a bus 10~20
Run time 2h
Route A predefined headway 3~10 min Route B predefined headway 3~10 min Route C predefined headway 3~10 min Route D predefined headway 3~10 min Route E predefined headway 3~10 min
26
5.2 Simulation Results
In the preliminary experiment, we fine the appropriate value of each weight (1:0.5,
2:0.5, 3:0.75) to reduce the headway deviation and schedule delay of buses and don’t produce huge impact on the passengers’ waiting time. And we use these value in the below simulation.
We first evaluate the waiting time of buses in ATCB and other control strategies, which is the sum of waiting time at intersection of all buses. Figure 11 is the result of total waiting of all buses, and it shows that our scheme ATCB has the best performance at reducing bus waiting time, and it is better than other methods at least 40 percent. Both TDTSP and ADT consider bus priority which extends its current phase, the former one considers headway deviation, and the later one considers both the ordinary vehicles and buses. They have better performance than the actuated scheme which doesn’t consider bus priority. The non-fixed phase scheme doesn’t consider bus priority, but the non-fixed scheme is better than actual due to it is better for all vehicles include buses and ordinary vehicles.
Figure 11. Total waiting time of buses
0 20 40 60 80 100 120
0 1 2 3 4 5 6 7 8 9 10x 104
Time (min) Total waiting time (sec) ATCBTDTSP
Actuated AFLC Non-fixed
27
Table 2. Total waiting time of buses and average waiting time of each bus in one intersection
Scheme Total waiting time in two hours
Average waiting time of each bus in one intersection
ATCB 3.79 × 104 sec 12.31 sec
TDTSP 7.04 × 104 sec 22.57 sec
Actuated 8.79 × 104 sec 28.04 sec
AFLC 6.71 × 104 sec 21.72 sec
Non-fixed 6.38 × 104 sec 20.66 sec
We could see that the best performance is non-fixed scheme, since that the non-fixed scheme don’t concern bus priority. And our ATCB concern bus priority from bus schedule delay, headway deviation, number of passengers so our total waiting time of all vehicles is more than the non-fixed scheme. But the difference is just 3.8 percent, and it is acceptable because we can save 40 percent waiting time of buses as shown in Figure 11. And TDTSP has the poor performance due to that it doesn’t concern the real-time information except bus information. Actuated scheme extends the current phase to permit vehicles pass through intersections is more better than TDTSP due to that it can avoid vehicles wasting a lot of time waiting for signals just because it misses few seconds to pass through the intersection.
28
Figure 12. Total waiting time of vehicles
Table 3. Total Waiting time of buses and average waiting time of each vehicle in one intersection
Scheme Total waiting time in two hours
Average waiting time of each vehicle in one intersection
ATCB 1.47 × 107 sec 25.95 sec
TDTSP 2.23 × 107 sec 38.11 sec
Actuated 1.55 × 107 sec 26.78 sec
AFLC 1.61 × 107 sec 27.26 sec
Non-fixed 1.42 × 107 sec 24.67 sec
The passenger’s total waiting time is shown in Figure 13. And we could find that the result shown in Figure 13 is similar to Figure 12, but the performance of our ATCB is better than the non-fixed phase scheme, because we consider the number of passenger of each vehicle in our equation (5). If we don’t consider number of passengers of each vehicles like non-fixed phase scheme, the bus will be treated as ordinary vehicle without
0 20 40 60 80 100 120
0 0.5 1 1.5 2 2.5x 107
Time (min) Total waiting time (sec) ATCBTDTSP
Actuated AFLC Non-fixed
29
bus priority. And we could see it has better performance of ordinary vehicles as shown in Figure 12, but it has poor performance in the result of passengers’ waiting time.
Figure 13. Total waiting time of passengers
Table 4. Total waiting time of passengers and Average waiting time of each vehicle in one intersection
Scheme Total waiting time in two hours
Average waiting time of each passenger in one intersection
ATCB 1.50 × 107 sec 24.32 sec
TDTSP 2.28 × 107 sec 36.96 sec
Actuated 1.61 × 107 sec 26.10 sec
AFLC 1.66 × 107 sec 26.91 sec
Non-fixed 1.46 × 107 sec 23.67 sec
Then we evaluate the schedule delay of buses. We need buses travel a distance to produce bus schedule delay, so we collect the information from pivot intersection, and this intersection is a congested intersection because it is passed through by buses of all bus routes. As shown in Figure 14 and Table 5, we record 30 records of the schedule
0 20 40 60 80 100 120
0 0.5 1 1.5 2 2.5x 107
Time (min) Total waiting time (sec) ATCBTDTSP
Actuated AFLC Non-fixed
30
delay of buses on RouteA when it arrives at this intersection. We could find that our ATCB has a lowest average schedule delay, which is better at least 25 percent than other schemes. The variance of bus schedule delay of ATCB is 337.79, and it is better than other schemes at least 63.7 percent.
Figure 14. Bus schedule delay in a bus stop at a pivot intersection
Table 5. Average schedule delay time and Variances of buses at a pivot intersection
Scheme Average schedule delay time Variances of bus schedule delay
ATCB 30.87 sec 337.79 sec
TDTSP 46.88 sec 930.52 sec
Actuated 47.77 sec 1360.84 sec
AFLC 57.22 sec 1971.85 sec
Non-fixed 55.80 sec 985.0 sec
5 10 15 20 25 30
0 20 40 60 80 100 120 140 160 180
Bus (veh)
Bus delay (sec)
ATCB TDTSP Actuated AFLC Non-fixed
31
As shown in Figure 15 and Table 5, we evaluate headway deviation of buses. We still collect records of bus headway deviation ratio at the pivot intersection. As shown in Figure 14 and Table 5, in ATCB, the average bus headway deviation ratio is 22.19 percent, and the range of bus headway deviation ratio is from -44 percent to 39 percent and variance is 601.54 percent. The second best performance is TDTSP, and its average bus headway deviation ratio is 22.19 percent, and the range of bus headway deviation ratio is from -68 percent to 70 percent and variance is 601.54 percent.
We also evaluate the total schedule delay time and bus headway deviation ratio which is shown in Table 7, and the result is similar to the result shown in Table5 and Table6.
Our ATCB still have the best performance.
Figure 15. Bus headway deviation ratio in a bus stop at a pivot intersection
5 10 15 20 25 30
-80 -60 -40 -20 0 20 40 60 80 100 120
Bus (veh)
Bus deviation ratio (%)
ATCB TDTSP Actuated AFLC Non-fixed
32
Table 6. Average headway deviation ratio and variances of buses at a pivot intersection
Scheme Average headway deviation ratio
Variances of bus headway deviation ratio Scheme Total waiting
time of
33
5.3 A Simulation in Real Urban Environment
To prove ATCB is still effective in real world, we have designed a road network of Taipei City in Taiwan, a rough map is shown in Figure 16. This network has 16
intersections. And we have select 5 real bus routes to simulate.
The result is shown in Table 8, and we can find that the result is similar to the performance in Table 7, that can prove our method ATCB is still effective in a real road network. But because the size of this map is small, so the benefits of reducing waiting of buses are more obvious than the benefits of buses headway and bus schedule than the result of the bigger map.
Shimin Blvd
Zhongxiao East Rd
Renai Rd
Xinyi Rd
Guangfu South Rd
Dunhua South Rd
Fuxing South Rd
Jianguo South Rd
Figure 16. Road network of Taipei City
34
Table 8. Simulation result of a real urban environment Scheme Total waiting
time of
35
Chapter 6 Conclusion
In this paper, the adaptive traffic signal control system (ATCB) has been proposed for ordinary vehicles and buses in urban environment. We adopt non-fixed phase scheme, and we use collected traffic information from detectors on roads and buses to determine the next phase and the phase should be allocated. To determine which phase is suitable to be adopted, we concern the passengers’ waiting time and bus priority which include bus headway deviation and bus schedule delay to determine the demand of each phase.
And the phase which cause lowest passengers’ waiting and improve schedule delay and headway deviation will be selected. The simulation results show that ATCB performs better at reducing passenger’s waiting time at least 40 percent, and improving schedule delay at least 17.5 percent, and headway deviation of buses at least 6 percent.
For the future works, the prediction model of traffic flow is also an important point in traffic signal control systems, and we could have better performance by implementing the prediction model of traffic flow in our system. Furthermore, we will try to use the information of neighbor intersections to design better control scheme.
36
References
.
[1] D. I. Robertson and R. D. Bretherton, “Optimizing networks of trafficsignals in real timełthe scoot method,” IEEE Transactions on Vehicular Technology,vol. 40, no. 3, pp. 11–15, 1991.
[2] A. G. Sims and K. W. Dobinson, “The sydney coordinated adaptive traffic (scat) system philosophy and benefits,” IEEE Transactions on Vehicle Technology, vol.
29, no. 2, pp. 130–137, 1980.
[3] E. Y. Luz and A. Mimbela “A summary of vehicle detection and surveillance technologies use in intelligent transportation systems.”
http://www.nmsu.edu/~traffic/
[4] B. Zhou, J. Cao, X. Zeng, and H. Wu, “Adaptive Traffic Light Control in Wireless Sensor Network-Based Intelligent Transportation System,” IEEE Conference on Vehicular Technology, pp1-5 May 2010.
[5] N. Maslekar, M. Boussedjra, J. Mouzna, and H. Labiod,“VANET Based Adaptive Traffic Signal Control.” IEEE Conference on Vehicular Technology, pp. 1-5, 2011.
[6] I. Arel, C. Liu, T. Urbanik, and A. Kohls, “Reinforcement learning-based multi-agent system for network traffic signal control,” IET Intelligent Transport Systems, vol. 4, no. 2, pp. 128–135, 2010.
[7] A. Guerrero-Ibanez, J. Contreras-Castillo, R. Buenrostro, A. B. Marti, and A. R.
Munoz, “A policy-based multi-agent management approach for intelligent
37
traffic-light control”, IEEE Intelligent Vehicles Symposium, pp. 694-696, June 2010.
[8] K. Mizuno, Y. Fukui, S. Nishihara, “Urban Traffic Signal Control Based on Distributed Constraint Satisfaction,” Hawaii International Conference on System Sciences, pp. 65, 2008.
[9] T. Li, D. Zhao, and J. Yi, “Adaptive Dynamic Programming for Multi-intersections Traffic Signal Intelligent Control,” IEEE Conference on Intelligent Transportation Systems, pp. 286-291, 2008.
[10] B. Zhou, J. Cao, H. Wu, “Adaptive Traffic Light Control of Multiple Intersections in WSN-based ITS”, IEEE Conference on Vehicular Technology, pp15-18 May 2011.
[11] T. H. Heung, T. K. Ho, and Y. F. Fung, “Coordinated road-junction traffic control by dynamic programming,” IEEE Transactions on Intelligent Transportation Systems, vol. 6, no. 3, pp. 341–350, Sep. 2005.
[12] P. B. Mirchandani and F.-Y. Wang, “RHODES to intelligent transportation systems,” IEEE Intell. Syst., vol. 20, no. 1, pp. 10–15, Jan./Feb. 2005.
[13] T. H. Heung, T. K. Ho, and Y. F. Fung, “Coordinated road-junction traffic control by dynamic programming,” IEEE Transactions on Intelligent Systems, vol. 6, no. 3, pp. 341–350, Sep. 2005.
[14] Y. W. Kim, T. Kato, S. Okuma, and T. Narikiyo, “Traffic network control based on hybrid dynamical system modeling and mixed integer nonlinear programming with convexity analysis,” IEEE Transactions on Systems, Man and Cybernetics, vol. 38, no. 2, pp. 346–357, Mar. 2008.
38
[15] Y. S. Murat and E. Gedizlioglu, “A fuzzy logic multi-phased signal control model for isolated junctions,” Transportation Research Part C: Emerging Technologies, pp.
19–36, 2005.
[16] D. Srinivasan, M. C. Choy, and R. L. Cheu, “Neural networks for realtime traffic signal control,” IEEE Transactions on Intelligent Systems., vol. 7, no. 3, pp. 261–272, Sep. 2006.
[17] J. D. Schmocker, S. Ahuja, and M. G. H. Bell, “Multi-objective signal control of urban junctions—Framework and a London case study,” Transportation Research Part C: Emerging Technologies, pp. 454–470, 2008.
[18] N. Hounsell, B. Shrestha, J. Head, S. Palmer, and T. Bowen, “The way ahead for London’s bus priority at traffic signals,” IET Intelligent Transport Systems, vol. 2, no. 3, pp. 193–200, Sep. 2008.
[19] W. Ma, X.Yang, “Design and Evaluatio Priority System Base on Wireless Sensor Network”, IEEE conference on Intelligent Transportation Systems, pp.
1073-1077,Sep2008.
[20] Bhouri, N.; Haciane, S.; Balbo, F., “A Multi-Agent System to Regulate Urban Traffic: Private Vehicles and Public Transport”, International IEEE conference on Intelligent Transportation Systems, pp1575-1581,2010.
[21] G. J. Shen and X. J. Kong, “Study on road network traffic coordination control technique with bus priority,” IEEE Transactions on Systems, Man and Cybernetics, vol. 39, no. 3, pp. 343–351, May 2009.
[22] Hounsell, N.; Shrestha, B., “A New Approach for Co-Operative Bus Priority at Traffic Signals,” IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 1, pp. 6–14, Sep. 2005.
39
[23] Wilensky, U. NetLogo (and NetLogo User Manual), Center for Connected Learning and Computer-Based Modeling, Northwestern University.
http://ccl.northwestern.edu/netlogo/, 1999.